II-D Encoding Positions The attention modules don't look at the get of processing by design. Transformer [62] released “positional encodings” to feed details about the position on the tokens in input sequences. LLMs demand extensive computing and memory for inference. Dep